In normal cells, DNA methylation is rare in promoter-associated CpG islands but important to X-inactivation, genomic imprinting and repression of repetitive elements. During tumorigenesis hundreds to thousands of genes gain methylation in promoter-associated CpG islands, affecting many pathways including tumor- suppressor pathways. The causes of this massive switch in DNA methylation remain mysterious. Interestingly, some genes are refractory to abnormal de-novo methylation in cancer while others are frequently targeted. Deciphering the differences between CpG islands sensitive or resistant to aberrant methylation will lead to a better understanding of the cellular factors that modulate cancer-specific DNA hypermethylation. Based on preliminary data, our central hypothesis to explain this difference is that an interplay between local sequence features (repeat elements, recognition sites for DNA binding proteins) and baseline chromatin states (histone modifications) related to developmental transcription programs modulates CpG island methylation in cancer. To test this hypothesis, we propose the following specific aims: (1) Identify baseline genetic and epigenetic features which segregate genes with propensity to become de-novo methylated in cancer from genes protected from de-novo methylation; and (2) use cellular models of DNA methylation induction to validate the individual and cooperative action of candidate genetic and epigenetic features. In specific aim 1, we propose (a) to measure the propensity of promoter-associated CpG islands to DNA hypermethylation in myeloid leukemia and colorectal carcinomas) and (b) to identify the genomic (transcription factor binding sites, retrotransposons, short direct repeats) and epigenomic (histone modifications in normal cells) factors that distinguish methylation-prone versus methylation-resistant CpG islands. Significant features will be entered in a mathematical model to reveal individual and cooperative activity in modulating methylation in cancer, and the model will be validated in other samples and tumor types. The most significant features and known factors associated with differential predisposition to DNA methylation (the transcription factor Sp1, LINE/SINE retrotransposons and the insulator proteins CTCF, USF1/2 and VEZF1) will be tested in specific aim 2. For this testing we will use a series of cellular models developed in our laboratory where engineered transgenes can be inserted in specific genomic loci, and later on moved in and out of repressive contexts due to the presence of tetracycline-induced repressors. We expect that the successful development of our research will bring novel insights in how abnormal DNA methylation is targeted to specific genes while sparing others, and will also result in the identification of multiple targets for epigenetic-based therapies.